Number entity Bugs


I created a simple Dialog task which has a Number entity.
The dialog goes as follows:

Intent node —> Entity node(Number entity) —> Message

When i give input as 10K , entity is not getting identified. The reply from RTM is that “I’m sorry, I did not recognize the number you entered. Please enter a number again.”

whereas if i give input as 10[space]K (10 K) entity is getting identified with 10,000.
Kore kore1

Hi @springer.etaeere,

We are able to replicate the scenario.
There is a difference on how the bot reads the number value 10K , 10 k, 10 K.

We will discuss this with our engineering team and get back to you with appropriate explanation and solution.

Yoga Ramya.

Why should the bot read 10K & 10 K differently? It seems that both the scenarios (With [space] and without space) are valid! I really wonder how can a single space make a difference!

Some similar examples:

  • 3M and 3 M
  • 4K and 4 K
  • 5G and 5 G

Not saying that 10K/ 10 K shouldn’t be handled the same way, but it is not always clear cut.

But why doesn’t it work the same way? Both seems to be legit in format right?
Is there any specific reasons why (with space) is not being handled in the bot?

There was a fairly recent issue reported by a customer about the currency entity not picking up 10 K as 10 MMK or 10 PGK, and was just seeing 10,000. It looks like that has been a small regression in addressing that.

The K symbol is either a kyat for the currency of Myanmar or a kina for the currency of Papua New Guinea.

I was able to identify the currency codes of myanmar and papua. And the bot functioning also seems to be legit according to your explanation.
So as of now, both 10K and 10 K should be finding PGK and MMK codes. But it is not able to due to regression issue (as you have said)

But why is number entity not able to identify k as 1000, M as million and G as grand when there is a space between a number and unit? As in the case of currency that 10K and 10 K should work the same way, even in number entity also it should function the same way right?

It is certainly the intent that all of these cases work should work as expected. 10K and 10 K should be able to be interpreted as 10000 or 10 kina or 10 Kelvin depending on the current entity that is being processing. It is just that that there was a small regression introduced recently in fine tuning the handling of 10K/10 K within a currency entity.

Note there is a complicated relationship between currency and number because the former does support just a number. In speech we often don’t include an explicit currency because a context is assumed, so if at a currency prompt it can be tricky to determine if 10K is implicitly 10000 or explicitly 10 kina.

It will be fixed.